=Paper=
{{Paper
|id=Vol-1920/paper3
|storemode=property
|title=Artifact-Centric Business Process Models in UML: Specification and Reasoning
|pdfUrl=https://ceur-ws.org/Vol-1920/paper3.pdf
|volume=Vol-1920
|authors=Montserrat Estañol
|dblpUrl=https://dblp.org/rec/conf/bpm/Estanol17
}}
==Artifact-Centric Business Process Models in UML: Specification and Reasoning==
https://ceur-ws.org/Vol-1920/paper3.pdf
Artifact-centric Business Process Models in UML:
Specification and Reasoning (Extended Abstract)
Montserrat Estañol
supervised by Prof. Ernest Teniente
Universitat Politècnica de Catalunya, Barcelona, Spain
{estanyol|teniente}@essi.upc.edu
1 Introduction
Processes can be modeled from different perspectives. Traditional process mod-
eling has followed the process-centric perspective, where the focus is on the
sequencing of activities (i.e. the control flow), largely ignoring the data required
by them. In contrast, the artifact-centric (or data-centric) process modeling ap-
proach focuses on defining the data required by the tasks or activities, and details
of the tasks themselves in terms of the changes they make to the data.
One of the challenges in this context is finding a way to represent business
processes from an artifact-centric perspective, in a way that is easy to understand
by the people involved in the business process and that is formal at the same
time. To achieve this we propose a framework, BAUML, which is based on using
a combination of UML and OCL models.
Once a process model has been defined, it is important to ensure its qual-
ity. This will avoid the propagation of errors to the process’s implementation.
Although there are many different quality criteria, we focus on the semantic
correctness of the model, answering questions such as does it represent reality
correctly? or are there any errors and contradictions in it?
Therefore, the second part of this thesis is concerned with finding a way to
determine the semantic correctness of our BAUML models. To begin with, we
propose a translation of a BAUML model into a DCDS (an external framework
to our work), to which model checking techniques can be applied to determine
the model’s correctness. However, DCDSs have been defined theoretically and
there is no tool that implements them.
To solve this, we created a prototype tool, AuRUS-BAUML, which is able
to translate our BAUML models into first-order logic and to reason on their
semantic correctness using an existing tool, SVTe. Another contribution is the
logic translation which is performed by the tool. Last but not least, we contribute
a study on the decidability of reasoning on the BAUML models.
To the best of our knowledge, this is the first approach using a combination
of UML and OCL models for modeling artifact-centric business processes which
also provides a way of checking their semantic correctness.
2 M. Estañol
2 Modeling Artifact-centric BPM in UML/OCL
The BALSA framework [3] defines four different dimensions that should be
present in any artifact-centric business process model. They are the following:
business artifacts, which represent meaningful data for the business, and have an
ID; lifecycles, which represent the evolution of an artifact; services, representing
tasks (i.e. meaningful units of work) that evolve the artifact; and associations,
representing constraints in the manner how services make changes to artifacts.
In traditional, process-centric models, such as a flowchart, there is usually
only one of the BALSA dimensions represented: the flow between the tasks in
the process, called associations in the framework. In terms of artifact-centric
approaches, there are several alternatives (see [2] for details). However, none of
the examined works has the following characteristics: 1. It uses a well-known,
formal language, for all the dimensions in BALSA, which can be understood
by business modelers and developers; 2. The language bridges the gap between
artifact and process-centric approaches; 3. There are tools to model the business
process following the approach.
The modeling approach that we propose to deal with these issues is to use
a combination of UML and OCL models. UML class diagrams for business ar-
tifacts; UML state machine diagrams for lifecycles; UML activity diagrams for
associations, and OCL operation contracts for services or tasks. We call our
approach BAUML (BALSA UML, for short).
<>
Bicycle
id : String
inServiceSince : Date
BicycleState {disjoint, complete}
Unusable Available InUse
unsusableSince : Date lastReturn : Date [0..1] expectedReturn : DateTime
0..1 0..1 0..1
BicycleRental
unusable bike is in {xor} is in startTime : DateTime
1 1 1
AnchorPoint User
number : Natural Blacklisted id : String
date : Date name : String
1..*
belongs to creditCard : Natural
validUntil : Date
1
Station
id : Natural
Fig. 1. Class diagram for the Bicing example, representing the business artifacts
We will illustrate the approach by means of an example based on a city bicycle
rental system (Bicing). Business artifacts correspond to some of the classes in
the UML class diagram (see Figure 1); they are those whose evolution results in
relevant states from the business’s point of view (Bicycle in our example). We
refer to the rest of classes as objects (e.g. User or AnchorPoint). Due to space
Artifact-centric BPM in UML (Extended Abstract) 3
limitations we do not include any integrity constraints, which should be defined
in OCL.
Register New Bicycle
Pick Up Bicycle [success]
Available Return Bicycle InUse
Pick Up Bicycle [fail]
Register New Bicycle
Repair Bicycle [success]
Assign to
Unusable Repair Bycicle [fail] Create New Bicycle
AnchorPoint
Fig. 2. Lifecycle of artifact Bicycle. Fig. 3. Act. diagr. of RegisterNewBicycle.
For each artifact, a state machine diagram will show its lifecycle (see Figure
2). Each of the states corresponds to one of the subclasses of Bicycle, and each
transition shows how a Bicycle evolves from one state to the next. Then, each
transition of the state machine diagram is further specified (as they are not
atomic) by means of an activity diagram determining the associations between
the services or tasks of the artifact. See the activity diagram of Register New
Bicycle in Figure 3. Finally, the behavior of the tasks from each activity diagram
is defined through an OCL operation contract, which contains a precondition,
stating the conditions that must be true in order for the operation to execute,
and a postcondition, indicating the state of the system after the execution of
the operation. Below we show the operation contract for task CreateNewBicycle,
which has no precondition and which creates a new bicycle.
operation c r e a t e N e w B i c y c l e ( bId : String ) : Bicycle
post : Available . allInstances () -> exists ( b | b . oclIsNew () and b . uid = bId and
b . inServ iceSin ce = today () and result = b ) )
The framework presented here has several advantages: it uses two languages,
UML and OCL, both of which are ISO standards. They allow us to define the
business process at a high level of abstraction and thus they are independent
of the final technological implementation. These languages, but more specially
UML, can be understood by business modelers and developers. In addition, our
proposal has precise semantics and can deal with business processes that contain
more than one artifact [2].
3 Reasoning on Artifact-centric Business Process Models
Given a BAUML model, our goal is to be able to ensure that it fulfills a certain
set of desirable properties. In particular, we wish to check that there are no
errors in the model and that it represents reality correctly. Unfortunately there
are no available methods to reason with models used in BAUML.
4 M. Estañol
Data-centric Dynamic Systems (DCDSs) are an alternative way of represent-
ing data-centric business process models. They provide a formal representation
at a lower level of abstraction than UML and OCL, but it is possible to ap-
ply model checking techniques to them in order to check if they fulfill a certain
property [1].
To take advantage of DCDSs, the thesis presents and formalizes (by means
of algorithms) a translation process to obtain a DCDS that is equivalent to a
BAUML model. A relational DCDS is a tuple S = hD, Pi, where D corresponds
to the data layer and P to the process layer. The data layer contains, among
others, a database schema R and an initial database instance I0 .The process
layer contains, among others, a set of actions A and a set of condition-action
rules of the form Q 7→ α, where Q is a first-order query and α ∈ A.
Intuitively, we could assume that one way of performing the translation pro-
cess from BAUML to DCDSs is by mapping the models according to the BALSA
dimension which they represent, e.g. by translating the class diagram into a
database schema R as both represent the business artifacts. However, the trans-
lation process requires additional considerations, such as adding tables to track
the evolution through the tasks or services in the activity diagram [2].
Once the model has been translated and the property has been defined, we are
able to apply model checking techniques to determine if the original BAUML
model fulfills said property. To do so, we will also need to provide an initial
instance of the database schema.
Unfortunately, there is no tool that can reason with DCDSs to check their
semantic correctness. For this reason, we have created a prototype tool, AuRUS-
BAUML, which is able to perform several semantic tests to check the correctness
of a BAUML model, by translating the starting models into first-order logic. The
translation process performed by the tool is a contribution of the thesis based
on the work of [4]. Note that this approach does not need an initial instance of
the model to obtain results, unlike DCDSs.
In particular, the tool can perform two types of semantic tests: verification
and validation. Verification ensures that there are no internal errors and con-
tradictions in the model (e.g. guaranteeing executability of the tasks), and can
be generated automatically; validation focuses on external correctness: ensuring
that the model fulfills the user requirements (e.g. checking that blacklisted users
cannot rent bicycles).
Another contribution of the thesis is a study on the decidability of reasoning
for the BAUML framework proposed in it. Determining whether an unrestricted
BAUML model fulfills a certain property is undecidable. By establishing certain
conditions over them, it is possible to ensure decidability over them without
bounding the number of active artifact instances. On the other hand, if the two
artifacts share read-write objects, we need to bound the number of active artifact
instances to ensure decidability. See [2] for details.
Artifact-centric BPM in UML (Extended Abstract) 5
4 Conclusions
The thesis presents a way to model business processes following an artifact-
centric approach using a combination of UML and OCL models, which results
in a high-level and mainly graphical representation of the process. We call this
approach the BAUML framework.
Since the models represent both the relevant data and how the tasks in the
processes make changes to this data, we can check the semantic correctness of the
models. This is a another contribution of the thesis. We propose two different
ways to do. The first one relies on translating the BAUML to a DCDS, an
external framework to our own work.
However, as DCDSs have been proposed theoretically, we then prove the
feasibility of our approach by implementing a prototype tool, AuRUS-BAUML,
which can translate the initial models into first-order logic automatically and
generate tests from them. It is also possible to execute user-defined tests. This
allows the detection of internal errors and contradictions (verification) and that
the model fulfills the requirements (validation).
Finally, a last contribution is a study on the decidability of reasoning on our
BAUML models. Determining whether a BAUML model fulfills a certain prop-
erty is undecidable, but we establish certain conditions to guarantee decidability.
As further work, we would like to incorporate other notations such as BPMN
or ER diagrams to make the framework more flexible. Another point to focus on
would be to address the limitations of the translation process or the tools (e.g.
by making it more efficient).
Acknowledgments: The work was partly supported by Universitat Politècnica
de Catalunya, Ministerio de Ciencia e Innovación under project TIN2011-24747,
Ministerio de Economía y Competitividad under project TIN2014-52938-C2-2-R
and AGAUR agency under project 2014 SGR 1534.
References
1. Bagheri Hariri, B., Calvanese, D., Giacomo, G.D., Deutsch, A., Montali, M.: Veri-
fication of relational data-centric dynamic systems with external services. In: Hull,
R., Fan, W. (eds.) PODS. pp. 163–174. ACM (2013)
2. Estañol, M.: Artifact-centric Business Process Models in UML: Specification and
Reasoning. Ph.D. thesis, Universitat Politècnica de Catalunya (2016), advised by
Prof. Ernest Teniente. Available at: http://hdl.handle.net/2117/96329
3. Hull, R.: Artifact-centric business process models: Brief survey of research results
and challenges. In: Meersman, R., Tari, Z. (eds.) OTM 2008. LNCS, vol. 5332, pp.
1152–1163. Springer (2008)
4. Queralt, A., Teniente, E.: Reasoning on UML conceptual schemas with operations.
In: van Eck, P., Gordijn, J., Wieringa, R. (eds.) CAiSE 2009. pp. 47–62. LNCS,
Springer (2009)